## HPL Frequently Asked Questions

- What problem size N should I run ?
- What block size NB should I use ?
- What process grid ratio P x Q should I use ?
- What about the one processor case ?
- Why so many options in HPL.dat ?
- Can HPL be outperformed ?

Navigation:

- › code/
- »
*downloads/* - › 9
- › RHEL6/
- › aplay/
- »
*mirrors/* - › ethan.uk.to/
- »
*netlib.org.benchmark/* - › bipopt/
- »
*hpl/* - › HPL abort
- › HPL all reduce
- › HPL barrier
- › HPL bcast
- › HPL binit
- › HPL broadcast
- › HPL bwait
- › HPL copyL
- › HPL daxpy
- › HPL dcopy
- › HPL dgemm
- › HPL dgemv
- › HPL dger
- › HPL dlacpy
- › HPL dlamch
- › HPL dlange
- › HPL dlaprnt
- › HPL dlaswp00N
- › HPL dlaswp01N
- › HPL dlaswp01T
- › HPL dlaswp02N
- › HPL dlaswp03N
- › HPL dlaswp03T
- › HPL dlaswp04N
- › HPL dlaswp04T
- › HPL dlaswp05N
- › HPL dlaswp05T
- › HPL dlaswp06N
- › HPL dlaswp06T
- › HPL dlaswp10N
- › HPL dlatcpy
- › HPL dlocmax
- › HPL dlocswpN
- › HPL dlocswpT
- › HPL dmatgen
- › HPL dscal
- › HPL dswap
- › HPL dtrsm
- › HPL dtrsv
- › HPL equil
- › HPL fprintf
- › HPL grid exit
- › HPL grid info
- › HPL grid init
- › HPL idamax
- › HPL indxg2l
- › HPL indxg2lp
- › HPL indxg2p
- › HPL indxl2g
- › HPL infog2l
- › HPL jumpit
- › HPL ladd
- › HPL lmul
- › HPL logsort
- › HPL max
- › HPL min
- › HPL numroc
- › HPL numrocI
- › HPL pabort
- › HPL packL
- › HPL pddriver
- › HPL pdfact
- › HPL pdgesv
- › HPL pdgesv0
- › HPL pdgesvK1
- › HPL pdgesvK2
- › HPL pdinfo
- › HPL pdlamch
- › HPL pdlange
- › HPL pdlaprnt
- › HPL pdlaswp00N
- › HPL pdlaswp00T
- › HPL pdlaswp01N
- › HPL pdlaswp01T
- › HPL pdmatgen
- › HPL pdmxswp
- › HPL pdpancrN
- › HPL pdpancrT
- › HPL pdpanel disp
- › HPL pdpanel free
- › HPL pdpanel init
- › HPL pdpanel new
- › HPL pdpanllN
- › HPL pdpanllT
- › HPL pdpanrlN
- › HPL pdpanrlT
- › HPL pdrpancrN
- › HPL pdrpancrT
- › HPL pdrpanllN
- › HPL pdrpanllT
- › HPL pdrpanrlN
- › HPL pdrpanrlT
- › HPL pdtest
- › HPL pdtrsv
- › HPL pdupdateNN
- › HPL pdupdateNT
- › HPL pdupdateTN
- › HPL pdupdateTT
- › HPL perm
- › HPL pipid
- › HPL plindx0
- › HPL plindx1
- › HPL plindx10
- › HPL pnum
- › HPL ptimer
- › HPL ptimer cputime
- › HPL ptimer walltime
- › HPL pwarn
- › HPL rand
- › HPL recv
- › HPL reduce
- › HPL rollN
- › HPL rollT
- › HPL sdrv
- › HPL send
- › HPL setran
- › HPL spreadN
- › HPL spreadT
- › HPL sum
- › HPL timer
- › HPL timer cputime
- › HPL timer walltime
- › HPL warn
- › HPL xjumpm
- › IBM README
- › algorithm
- › copyright
- › documentation
- › errata
- »
*faqs* - › links
- › references
- › results
- › scalability
- › software
- › tuning

- › sparsebench/
- › top500
- › top500/

- › netlib.org.research/
- › pi/
- › r 36.net/
- › software.x11/
- › wolfgang/
- › www.9paste.net/
- › www.in ulm.de/
- › www.tuhs.org/

- › vga truetype
- › xfce4 mixer slackware/

- › images/
- › information/
- › paste/

- What problem size N should I run ?
- What block size NB should I use ?
- What process grid ratio P x Q should I use ?
- What about the one processor case ?
- Why so many options in HPL.dat ?
- Can HPL be outperformed ?